European Language Grid Python SDK

The European Language Grid is the primary platform for Language Technology in Europe. With the ELG Python SDK, you can use LT services and search the catalog inside your Python projects.

Installation

Via pip / PyPI

pip install elg

Register on the ELG

Please visit the ELG website to create a user account if you haven't got one already.

Functionalities

Use LTs

Initialize LT service

LT Services can be initialized as Service objects using the service_id parameter and an integer corresponding to the LT service you want to use.
In the following code example, 474 corresponds to the Cogito Discover Named Entity Recognizer (notice the ID at the end of the URL). For more LT services, search the catalog via the Catalog functionality (see section below) or visit the list of LT services on the ELG.

from elg import Service

# Init LT service using its ID
lt = Service.from_id(474)

This requires you to login to your ELG account via the URL that is printed on your terminal.

After successful login, your tokens are saved as ~/.cache/elg/tokens.json, so you do not need to log in again for subsequent calls.

Run LT service

You can either pass an input file or a string / raw text.

# Pass an input file that should be processed by the LT service
result = lt("path/to/file")

# You can also directly pass raw text to the LT service in most cases
result = lt("Did Nikola Tesla live in Berlin?")

Use corpus

from elg import Corpus

corpus = Corpus.from_id(913)
corpus.download()

Use the catalog

from elg import Catalog

catalog = Catalog()

# Search and get the result as a list of Entity
results = catalog.search(
    resource = "Tool/Service", # "Corpus", "Lexical/Conceptual resource" or "Language description"
    function = "Machine Translation", # function should be pass only if resource is set to "Tool/Service"
    languages = ["en", "fr"], # string or list if multiple languages
    limit = 100,
)

# search interactively 
catalog.interactive_search(
    search = "keyword1 keyword2 ...",
    resource = "Tool/Service", # "Corpus", "Lexical/Conceptual resource" or "Language description"
    function = "Machine Translation", # function should be pass only if resource is set to "Tool/Service"
    languages = ["en", "fr"], # string or list if multiple languages
)

Create a LT service object from the results

service = Service.from_entity(results[0])
result = service("Did Nikola Tesla live in Berlin?")

Get info of an entity

from elg import Entity

entity = Entity.from_id(476)
print(entity)

Benchmark

You can also run a benchmark that evaluates multiple services receiving the same input:

from elg import Benchmark

ben = Benchmark.from_ids([610, 624])

result = ben(["Bush is the president of the USA and lives in Washington", "My name is Rémi and I live in France"], number_of_runs=4)

df = result.compare()
print("General comparison:\n", df)
df = result.compare_results()
print("Comparison of the results:\n", df)
df = result.compare_response_times()
print("Comparison of the response time:\n", df)

You can investigate the results by saving the output from the bench call to a variable or by accessing bench.services.

CLI

elg-cli search *term1* *term2* ... *termN* --lang "" --resource "" --function ""
elg-cli info *id*
elg-cli run *id* --authentication_file path/to/tokens.json --data_file path/to/file